Coding Agent Index v2.0 ranks Claude Opus 5.5 in Claude Code first at 61.7

Artificial Analysis launched Coding Agent Index v2.0, evaluating models and agent harnesses across 290 tasks on three equally weighted benchmarks. Claude Code with Claude Opus 5.5 (max) leads overall at 61.7. The update replaces DeepSWE with Cognition's FrontierCode v1.1 and revises SWE-Atlas-QnA grading. Artificial Analysis says the mean score for 21 configurations evaluated on both versions fell from 54.8 to 47.4; the versions use different evaluation setups.